Fig. 6: Sensitivity analysis on the effect of sizes of alignment and reference sets on the performance of the alignment method (scanner B).

In the sensitivity analysis on the size of the alignment set (top), we used the full reference set (3221 cases). Results are reported over 500 bootstrap samples. For the alignment size analysis, each bootstrap sample is created by sampling one alignment set of the size of interest from all available cases as well as one evaluation set (n = 2500 cases). For the reference size analysis, each bootstrap sample is created by sampling one reference set of the size of interest from all available cases as well as one evaluation set (n = 2500 cases). On the left, the points depict the average SEN/SPC over samples and error bars represent the 95% bootstrap confidence interval. On the right, each box shows the 25%, 50% and 75% percentiles of the bootstrap distribution; whiskers denote the 5% and 95% percentiles and any point outside of this range is represented as an outlier. This analysis shows that with as few as 250 cases in the alignment set, we already get very good results, stable across repeated sampling experiments. In the analysis below, we measured the performance of methods for different reference set sizes, we used 500 cases for alignment and varied the size of the reference set. We can see that the method is not too sensitive to the size of the reference, even if the more data the reference distribution comprises, the better the results get. Source data are provided as a Source Data file.